Regression from D131400: cross-language LTO causes a crash in the
compiler on the NULL deref of Scope in `isa` call when Rust IR is
involved. Presumably, this might affect other languages too, and
even Rust itself without cross-language LTO when the Rust compiler
switched to LLVM 16.

Reviewed By: rnk

Differential Revision: https://reviews.llvm.org/D134616

Fix the use_lld() to use llvm_shlib_dir similarly to how use_clang()
does it.  This fixes use_lld() wrongly prepending llvm_libs_dir,
i.e. the directory with system-installed LLVM libraries before
the build directory of standalone build.  As a result, the shared
libraries from an earlier version of clang end up being used instead of
the newly built version when running the test suite prior to installing.

To reproduce the problem, build and install LLVM with dylibs first,
e.g.:

    cmake ../llvm -G Ninja -DCMAKE_BUILD_TYPE=MinSizeRel \
      -DCMAKE_INSTALL_PREFIX="${HOME}"/llvm-test \
      -DLLVM_BUILD_LLVM_DYLIB=ON -DLLVM_LINK_LLVM_DYLIB=ON \
      -DLLVM_INSTALL_UTILS=ON
    ninja install

Then build clang against that installation and run tests:

    export LD_LIBRARY_PATH=~/llvm-test/lib
    export PATh=~/llvm-test/bin:"${PATH}"
    cmake ../clang -G Ninja -DCMAKE_BUILD_TYPE=MinSizeRel \
      -DCMAKE_INSTALL_PREFIX="${HOME}"/llvm-test \
      -DCLANG_LINK_CLANG_DYLIB=ON -DLLVM_BUILD_TESTS=ON \
      -DLLVM_EXTERNAL_LIT="${PWD}"/bin/llvm-lit
    ninja check-clang

The tests will be run with LD_LIBRARY_PATH of:

    /home/${USER}/llvm-test/lib:/home/${USER}/llvm-project/build-clang/lib

As a result, installed libclang-cpp will take precedence over the one
from build dir.  With the patch, the correct path is used, i.e.:

    /home/${USER}/llvm-project/build-clang/lib:/home/${USER}/llvm-test/lib

Differential Revision: https://reviews.llvm.org/D135368

Differential Revision: https://reviews.llvm.org/D134917

Android does not use a shared libunwind.so and its presence in the build
directory can be harmful, for example it can end up being dynamically
linked against while not being present on the device.

Do not specify the execution directory in the remote execution script command line
for the compiler-rt builtin library tests. There is a single execution file tests
within the single directory. No need to pack all of them every time, just run one by one.

Allowed a single file execution without the execution directory.

Now only DXILTranslateMetadata uses DXILResources, so DXILResourceWrapper is only used by DXILTranslateMetadata.
Once we add lower for createHandle, DXILResourceWrapper will be used in more passes.
Also we can add resource index allocation in DXILResourceWrapper.

Reviewed By: beanz

Differential Revision: https://reviews.llvm.org/D135190

TARGET dummy arguments have fir.target attribute attached to them,
but globals do not have any sign of TARGET. This patch adds
target attribute for globals, which can be queried via
::fir::GlobalOp::getTarget().

Differential Revision: https://reviews.llvm.org/D135313

The wrong function overload was called.

Differential Revision: https://reviews.llvm.org/D135342

Summary:
This added a check for no unknown argument warnings. This apparently
occurs in the Windows toolchain as it cannot find a toolchain. This
patch fixes it by just ignoring this warning.

Using '|&' syntax for piping both stdout and stderr is not supported by older bash. macOS pre-installs 3.2.57 as of today, and it causes test failure due to unexpected token '&'. We can use '2>&1' instead to make it compatible as much as possible.

```
******************** TEST 'MLIR :: mlir-cpu-runner/invalid.mlir' FAILED ********************
Script:
--
: 'RUN: at line 1';   not mlir-cpu-runner --no-implicit-module llvm-project/mlir/test/mlir-cpu-runner/invalid.mlir |& llvm-project/build/bin/FileCheck llvm-project/mlir/test/mlir-cpu-runner/invalid.mlir
--
Exit Code: 2

Command Output (stderr):
--
mlir-cpu-runner/Output/invalid.mlir.script: line 1: syntax error near unexpected token `&'
/mlir-cpu-runner/Output/invalid.mlir.script: line 1: `set -o pipefail;{ : 'RUN: at line 1';   not mlir-cpu-runner --no-implicit-module llvm-project/mlir/test/mlir-cpu-runner/invalid.mlir |& llvm-project/build/bin/FileCheck llvm-project/mlir/test/mlir-cpu-runner/invalid.mlir; }'
```

Reference: https://stackoverflow.com/questions/35384999/what-does-mean-in-bash

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D135347

https://reviews.llvm.org/D134668 removed all `--` (double dashes) when using `plugin-opt` to pass linker options and replaced them with `-`. https://reviews.llvm.org/D133092 was committed later but introduced an instance of `--`. This patch replaces the `--` with `-`.

Reviewed By: MaskRay

Differential Revision: https://reviews.llvm.org/D135400

caused by 92233159

The `cl-uniform-work-group` attribute asserts that the global work-size
be a multiple of the work-group specified work group size. This should
allow optimizations. It is already present by default in the AMD
compiler and for HIP kernels so it should be safe to allow this for
OpenMP kernels by default.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D135374

This is a follow-up to D134224. The original patch added new `ExcludedHeader` enumerator to `ModuleMap::ModuleHeaderRole` and started associating headers with the modules they were excluded from. This was necessary to consider their module maps as "affecting" in certain situations and in turn serialize them into the PCM.

The association of the header and module needs to be handled when deserializing the PCM as well, though. This patch fixes a potential assertion failure and a regression. This essentially reverts parts of feb54b6d.

Reviewed By: Bigcheese

Differential Revision: https://reviews.llvm.org/D135381

The offloading toolchain makes heavy use of options beginning with
`--o`. This is problematic when combined with the joined `-o` flag. In
the following situation, the user will not get the expected output and
will not notice as the expected output will still be written.
```
clang++ -x cuda foo.cu -offload-arch=sm_80 -o foo
```

This patch introduces a warning that checks for joined `-o` arguments
that would also be a valid driver argument if an additional `-` were
added. I believe this situation is uncommon enough to warrant a warning,
and can be trivially fixed by the end user by using the more common
separate form instead.

Reviewed By: tra, MaskRay

Differential Revision: https://reviews.llvm.org/D135389

I accidentally cleared ForwardRefDSOLocalEquivalentNames twice instead.

Differential Revision: https://reviews.llvm.org/D135315

I tend to think we should ignore the policy bit in vsetvli insertion
if the tied operand is IMPLICIT_DEF. But that raises questions about
what the policy operand on RVV intrinsics means if you also pass
vundefined().

This change at least fixes some cases. I'll post a separate patch
for vsetvli insertion for discussion.

Reviewed By: reames

Differential Revision: https://reviews.llvm.org/D135386

[SmallVector] Reallocate if assigned memory is right after the current vector, created with capacity 0

Potential solution for
https://github.com/llvm/llvm-project/issues/57324.

Differential Revision: https://reviews.llvm.org/D132512

Signal is currently fullbuild only, so all functions inside it should be
as well.

Differential Revision: https://reviews.llvm.org/D135410

Add the kill syscall wrapper and tests.

Reviewed By: sivachandra

Differential Revision: https://reviews.llvm.org/D135406

Don't return a hard .FALSE. when folding IS_CONTIGUOUS() unless the
subscript triplets are discontiguous and array section is nonempty,
or if the array section is nonempty and there is a vector subscript.

Differential Revision: https://reviews.llvm.org/D135334

A namelist input item that is a derived type component reference
needs additional processing when the base item or the component
is an array.  When both have rank > 0, the component reference
must of course be subscripted.

(Fixes https://gitlab-master.nvidia.com/fortran/f18-stage/-/issues/999, NAG test t/tz2.)

I forgot to add documentation for these options when I added them to the `show` command, so add them now.

Reviewed By: phosek

Differential Revision: https://reviews.llvm.org/D135383

Previously the futex type was defined in terms of unsigned int, now it's
uint32, which is more portable.

Reviewed By: sivachandra

Differential Revision: https://reviews.llvm.org/D135408

We can lower these as an or with the negative of the condition value. This appears to result in significantly less branch-y code on multiple common idioms (as seen in tests).

Differential Revision: https://reviews.llvm.org/D135316

This is a non-trivial property relied upon by D135396. I wrote this to convince myself it was true.

Differential Revision: https://reviews.llvm.org/D135403

After generated call for ctor/dtor for entry, global variable for ctor/dtor are useless.
Remove them for non-lib profiles.
Lib profile still need these in case export function used the global variable which require ctor/dtor.

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D133993

This reverts commit d96ade00.

implement `std::ranges::basic_istream_view` and `std::views::istream`. Although the view itself is constexpr,
the constructor argument is a base class std::istream where its ctor/dtor are not constexpr. So no tests are performed in
constexpr

Differential Revision: https://reviews.llvm.org/D133317

A namelist input item that is a derived type component reference
needs additional processing when the base item or the component
is an array.  When both have rank > 0, the component reference
must of course be subscripted.

Differential Revision: https://reviews.llvm.org/D135218

This patch had to be reverted because on gcc 7.5.0 we see an error converting from std::unique_ptr<MCRegisterInfo> to Expected<std::unique_ptr<MCRegisterInfo>> as the return type for the function createRegInfo. This has now been fixed.

An error message in semantic analysis of SELECT TYPE type guards
doesn't allow for the possibility of failure from FindSourceLocation(),
and that can happen when the type guard is just an intrinsic type
name like "type is (character)".  Deal with it.

Differential Revision: https://reviews.llvm.org/D135217

Register allocation is split into two passes, and the expected behavior is that the first pass only should only work on virtual SGPRs. Whereas the second pass works on virtual VGPRs. This adds a test case which breaks if the first pass allocates VGPRs.

Differential Revision: https://reviews.llvm.org/D135331

This new pass for llvm-reduce attempts to reduce DebugInfo metadata.
The process used is:
  1. Scan every MD node, keeping track of nodes already visited.
  2. Look for DebugInfo nodes, then record any operands that are lists.
  3. Bisect though all the elements of the collected lists.

Differential Revision: https://reviews.llvm.org/D132077